Acerca de
"I don't know where to start."
This is probably the biggest excuse you give when starting to learn robotics and artificial intelligence.
In a certain sense, it is not really an excuse. It is perhaps even valid!
In today's day and age there is an abundance of resources made available to each and everyone of us. On the click of a button, you can get access to the best video lectures from the top universities in the world. You get access to novel projects and repositories contributed by great content creators and robotics and machine learning engineers and researchers. A variety of different types of courses are available to each and every one of us at a very nominal cost.
Having looked at this information, it does sound like you are making an excuse. If you were motivated enough, you should have been able to take advantage of the resources and start learning robotics, computer vision or machine learning already, right?
Although the vast amount of resources makes education accessible to everyone with a decent internet connection, resource abundance is a double-edged sword. Accessibility comes at the cost of overwhelm. Beginners with little-to-no knowledge about robotics find themselves easily overwhelmed as they do not know which resources to refer to. They try out a few arduino projects, take up an online course and watch a few tutorials online, never really finding one specific sub-domain they would want to master to a professional level.
If you just want to learn robotics and machine learning to build some fun side projects, it probably does not matter that much. You can take up any course that interests you and start doing any project that you have time, budget and motivation for.
However, if you are serious about learning robotics and/or AI and want to work professionally as a robotics software engineer, machine learning engineer, data scientist or computer vision engineer, telling you to buy a robotic kit or start tinkering with arduino would be a poor advise.
Arduinos and robotic kits are for hobbyists! if you want to work as a professional software engineer/ researcher in the robotics/ AI domain, you need a broad knowledge of several sub-domains in robotics and AI and deep expertise in your niche. And you need to have a system to be able to explore different fields such that you can choose a niche in which you want to gain expertise.
And, I can absolutely relate with you when you say, "I don't know where to start." In fact, when I was working as a mechanical design engineer and wanted to transition to robotics, I made the same excuse. I tried quite a few courses and was certainly making progress. But I lacked direction. And there was no guarantee that I would be able to work professionally in the field of robotics.
For me the only option was going for a Master's in Robotic systems engineering. I did not know any better.
The number 1 thing that I got from the master's programme was the well-defined structure that the University offered. The programme allowed me to explore several sub-domains in robotics and take up elective subjects of my interest. This was good exposure!
However, I was only able to cultivate the right skill-set that the industry would value by working for the industry. I took up multiple working student, and internship roles and additional projects in research institutes to gain exposure to enough sub-domains in robotics so that I could choose one specific niche, I would like to work in. And this is where a majority of my "real" learning was - Working for startups, bigger companies and research institutes on real-world projects.
And I understand that going for a Master's is not a real possibility for a large proportion of engineering students/ working professionals. There can be time constraints, monetary constraints, social and famial constraints, just to name a few.
I am here to tell you that it is absolutely possible to become a self-taught robotics software engineer/ machine learning or computer vision engineer. I can say this so confidently because of the following reasons -
-
Realize that M.Sc./ M.Tech in Robotics/ Data Science/ Machine Learning are just starting to become mainstream. This was not widely taught as a specialization even 5 years ago. There were fewer such courses. But the industries still employed machine learning engineers. Most of the people had to learn it on the job because one or more projects at their company required that.
-
There are countless examples of people who switched from mechanical/ electronics engineering to software engineering. Software engineering is at the core of AI.
-
For a new job, you have to learn new skills anyway. Even if you have prior experience in, let's say, computer vision, you have to deal with certain tools and frameworks that are specific to the company. So, you are learning on-the-job in any case.
-
A candidate who demonstrates their skill-set acquired by doing projects on their own shows self-initiative and that they have the ability to work independently and are disciplined enough. Such a candidate will often even stand out in comparison to other candidates.
-
At the end of the day, a company values the right skill-set in a candidate. Such a skill-set can be developed by taking part in practical projects. If you have a portfolio of projects that you developed, that is demonstration of a work experience. And such experience is in no way less as compares to someone who has a Master's in robotics/ data science. Your skill-set is your most valuable asset.
So, get rid of the idea that it is not possible to become a self-taught professional roboticist/ AI engineer. You just need to get rid of this limiting belief and start learning.
And, the single biggest reason why you constantly fail to start learning is because you lack a
clear, systematic, step-by-step, practical, project-based, personalized and well-defined Roadmap!
That's it. That's all you need. You need a system - A system that is tailored to your own individual goals and needs, is project-based and gives you the skill-set that would make you employable in the industry.
I have been been thinking about the problem of finding the "best way to learn robotics and AI" for a very long-time now. I started thinking about it when I was working full-time and started learning robotics on my own with only the internet at my disposal. Throughout my Master's programme, I was taking the good elements of the programme and tailoring those towards my own personal goals, finding my own area of interest that I would want to work professionally in. All the internships, working student positions and additional projects that I took were solely to gain enough exposure to understand the vast field of AI and robotics on a broad enough scale to be able to choose a niche, that would not just be of my interest but would also offer enough job opportunities.
I had the following realizations about what an ideal robotics curriculum should look like -
-
A good robotics curriculum should be project-based
-
A robotics programme should give enough exposure to the overall field to allow students to explore and find their niche
-
It should offer students both industry and research-exposure.
-
It should be individualized and tailored to the specific needs of the person. It should be curiosity-driven and personalized.
-
A good curriculum should also offer you accountability in terms of the right peer-group
-
Employability and practicality of the course are extremely important, especially if the goal is to get a job by means of the course.
-
A curriculum or a roadmap should also be dynamic in the sense that it should be updatable based on the recent advances in robotics and AI and not stay static year after year.
-
And to actually work, it needs to be engaging enough to the students.
After years of conceptualization of the idea, making use of my practical exposure to industry and academia, experience of studying in a top-notch conventional Master's programme, and just a lot of trial-and-error, I would like to share with you the exact roadmap that you can use to learn and master any sub-domain in robotics and AI.
The entire roadmap is covered in this video and if you follow the video till the end, and stick with the plan laid out to you at the end of the video, there is no reason why you cannot become a self-taught robotics software/ machine learning/ computer vision engineer.
Additionally, I plan to replicate an even more expanded version of this Roadmap by the end of the year 2023 in the form of a course. The course would contain the following modules -
Each module would have multiple activities that would personalized activities tailored towards your specific needs, goals, and interests.
An example of one such activity is demonstrated below.
To the best of my knowledge, no such personalized project-based curriculum, system or framework exists. And I would be happy to create such an in-depth course if the demand is high enough.
The course: